A Study on People Counting Through CSI

• Main Authors: PING-FENG CHEN, 陳屏鳳
• Other Authors: Hsu-Sheng Tsai
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/17920278489685723078

碩士 === 樹德科技大學 === 資訊管理系碩士班 === 104 === In recent years, due to the popularity of the internet and the development of wireless communication technology, most of the internet-related applications have made great progress. Currently the data collection for people counting issue relied on infrared imaging or by hands. In this research, the CSI (Channel State Information) collected from Wi-Fi 802.11n was adopted. By using MIMO (Multiple Input Multiple Output), objects with different sizes or even complex shapes are more likely to be detected. In this thesis, we first applied Intel 5300 NIC to collect Wi-Fi CSI data for people counting. Secondly, we use Octave to transform raw CSI data into numbers in matrix form for further statistically computation, such as average and SD variance analysis. The experiments were under six scenarios from no one to five persons in order to reveal their difference. In the experiment of adjacent walking, AP mode and non-AP mode were set respectively. Both modes show that SD increases with the number of people. However in non-AP mode is more obvious. In the experiment of people walking back and forth between sender and receiver, results also show that SD increases with the number of people. But, in non-AP mode is more obvious. We thus argue the influence of walking direction. In the experiment of people halting side by side or forming a circle around, the results show that SD changes slightly and is unable to differentiate. To sum up, human body is sensible to CSI, that is, by statistically analyzing CSI, the number of people may be estimated.